Metaverse Ethics, Innovation, New AI and the Circular Digital Economy
Dimitrios Kalogeropoulos, PhD
CEO Global Health & Digital Innovation Foundation | UCL GBSH Health Executive in Residence | EU AI Office GPAI Code of Practice | PhD AI in Medicine | IEEE European Public Policy Committee | Chair IEEE P3493.1? | Speaker
Retooling digital development
Over the past few years technology democratisation and the enablement of grassroots innovation have brought about the opportunity for more systemic and disruptive innovation, for transformational growth across sectors, for diversification and productivity improvements, and together the opportunity to regenerate revenue from shared data spaces.
Opportunities have in turn been driving the acceleration of economic activity digitalisation - digital development at large and at global scale, now reaching rates which by far outperform those of non-digital GDPs and promising to deliver further productivity gains, more data generated revenue, new business models and markets, and the complete transformation of value-added services. In health, the majority of this transformational growth is data-driven and hence any digital future shall depend upon our ability to share data in order to coproduce better services. Recycling data is a natural part of such a future.
In the digital economy there is plenty of opportunity to go around: albeit measuring the digital economy is no easy task (1), it is estimated that ROI in terms of GDP produced in the US with digital investments runs at almost seven times that of non-digital investments. And as far as forecasts go, under a sustainable high-digitalization scenario, the global digital economy is set to grow to a quarter of global GDP by 2025.
Sustaining a high digitalization rate means sustaining the capacity for growth, capitalisation on spillovers and reinvestment, and recycling, all of which in turn depend on economies being able to a) Maintain an aggressive pace of digital investment, b) Work together to deliver a strong digital strategy, including i) a supportive infrastructure, ii) a thriving entrepreneurial class with an innovative financing environment, iii) the reversal of the slowdown in technology-driven productivity growth, and iv) an overall vibrant technology sector, c) Invest in upskilling and reskilling the workforce, d) Establish an adequate regulatory framework and mechanisms, and e) Ensure a stable and enfranchised social environment, f) a favourable institutional environment and g) responsive policies that broaden access to opportunities,?and g) Avoid overregulation or deregulation (2, 3).
Quite a list of preconditions that need to be fulfilled in order to maintain innovation momentum and economic growth. Moreover there is a catch: as is evident from a quick scan of the list, these preconditions lead to a Circular Dependency Predicament: that is, either directly or indirectly they depend on each other to perform properly.
Together with the Innovator's Predicament, concept defined further in this article, there is considerable risk that growth may become distorted, hindering economic dev-elopment and causing economic downturn. This is a concern particularly in the health sector.
"Shooting random people with blow-darts filled with the Johnson & Johnson vaccine in order to end the pandemic".
Hesitation to pursue some of the key preconditions for transformational growth in the health sector was temporarily set aside on account of the Covid-19 pandemic, as is the case with accelerated vaccine research, development and deployment (4,?5,?6, 7, 8). This acceleration will hopefully persist and some of the measures taken shall make it into the regulatory process together with other formally adopted governance frameworks for digital health (toolbox-A).
In addition, we need to strengthen and improve data health with data sharing, ethics, quality, reliability and trust (toolbox-B) - the very own DNA chain of innovation ecosystems and key ingredients for innovation enablement and platforming, all of which dependent on the above mentioned preconditions being met, and vice versa. To use our earlier innovation acceleration and deregulation example, data was a major impediment during Covid-19 vaccine deployment, as little of it was readily available in the proper format in order to evaluate the clinical effectiveness of vaccines and to accordingly adjust vaccination and public health policy (9).
Data sharing is impeded by a generalised lack of trust and trust can be reclaimed solely on the basis of the self-sovereign data governance paradigm (toolbox-A) associated technologies such as decentralised identification, trust services, blockchain and distributed ledger technologies (toolbox-C). In fact the whole approach to data sharing must be redesigned and re-regulated beyond the basic control over privacy offered by current frameworks (toolbox-A). For starters, there is a strong monopolisation trend to curb (10), considering the rate of growth attributed to the capitalisation achieved by technology giants (Apple, Google) and global platforms (Amazon, Netflix, Booking, Uber), and the data privacy and sovereignty issues associated with their terms of business.
Then there are the aforementioned objective difficulties in mapping out and hence supporting a broader digital development and growth. It is expected that a strategic framework for enablement, investment, regulation, and policy alignment shall facilitate the production of a detailed sector development map and the identification of further support needs. In this context we must consider the matter of balancing disruptive versus sustaining innovation investments (toolbox-D). And sustaining innovation, as illustrated below, is a matter of data sharing and trust building strategy, policy, regulation. Furthermore, it is only through data sharing that pain points in the system can be identified by policy makers and the market alike and innovation financed and supported by both sides.
In terms of digital development and innovation in health, the Innovator's Predicament is expressed as the inability of the innovator to view an experiment being pursued or the scaling of this experiment in terms of its integration into clinical, public health and health system practice and economies, outside their own vantage point. They can only perceive and factor-in the part of the model, platform and ecosystem which is directly relevant to their innovation and this is inherently irrelevant in clinical medicine and clinical research, certainly more so in public health, as these are inference arts and sciences which depend on denominators, continuity and integration to yield results and benefits to patients.
In medical AI one of the problems caused by this predicament, or one of the manifestations of the limitations caused, is bias exacerbation or the lack of longitudinal data which offer enough temporal evidence to render clinical trials reliable.
Overall, there is ample reason to strengthen both market and government efforts to expand, develop and diversify their respective capitalisation base for digital development. There is considerable opportunity for connected innovation and increasing returns to scale and scope. Then there is the objective to sustain growth. By capitalizing on regulated and connected digital development markets, governments and markets alike can support upscaling, robust and inclusive growth, and continued, sustainable, scaled-up and deconcentrated, demonopolized data-driven innovation. With that comes a global GDP share of 23 trillion by 2025.
Specifically, in the health sector, where most digital transformation and digital development are data-driven, regulation and governance are of paramount importance, to deliver a digitalisation and wider sector development environment which is sustainable and inclusive, which enables development activity that is deconcentrated and demonopolized, and which safeguards ethical development, while fostering diversity and connected innovation.
By democratising the global data sharing ecosystem, opportunity may be rebalanced in the health sector, and with enough assistance, capitalisation on digital development can also be democratised and demonopolized, and the health-sector-wide innovation ecosystem rebalanced. This article touches upon this key need and precondition for diversification, rebalancing and democratisation in data capitalisation in the health sector, provides a normalised view of the problem space, and offers a framework for development and capitalisation going forward.
Key points
Getting from A to B
Artificial Intelligence (AI) is the most talked about innovation and transformational growth platform for health, but also a pivotal aspect for the future tense on digital health. The latter because AI gives purpose to digital health - it raises the bar of expectations beyond mundane physician notes and patient records and brings us closer to immersive realities, virtual collaboration and health Metaverses. Quite the leap of expectations, but the pandemic clearly had this effect on the global innovation ecosystem. It seems though, with the exception of the regulatory aspects touched upon earlier, this was the only effect on the ecosystem, up until now at least.
In fact, AI has been, and for the foreseeable future shall continue to be, a very effective incentive to change, in expectations and the ecosystem too, as AI pre-empts digital development in order to make room for its assimilation; and this is important in a field which by far lags behind other sectors in digital development.
There are nonetheless two key underdeveloped areas in digital health. One key issue is how do we go from our current state A to a future state B of expectations. The other is defining what exactly A and B are.
As far as the latter is concerned, for instance, we should ask ourselves, have we, as a matter of fact, reached the capacity to productively deploy electronic health records (EHRs) across the care continuum in order to gather longitudinal (11) data on ground truths for our AI devices (12), among other destination states, or have we merely renamed electronic medical records (EMRs) into Big data in order to jump that hurdle?
As far as the transition goes, persisting fragmentation seems to have gotten the better part of the market, with little risk mitigation in place. In fact what is urgently needed is a high-quality moral compass, to provide for an organisational strategy that will point from A to B, and for routine policies that will guide us from A to B and will also help in defining A and B in a manner which is collective and mutually agreeable, hence normative and formative. With that, the ethics of digital development and AI are equated with governance and ethics frameworks with governance frameworks (13).
Moreover, our destination State-B becomes an ethical state and getting from A to B a transition from our current state of big data among space debris (junk) to an ordered metaverse of moral values, ethical data, and interconnected, productive communities, innovation ecosystems and economies.
In this destination reality, digital health is the sum of ideas and goods that produce innovation in health and improve the quality, efficacy, safety and outcomes from health services, ensuring equitable access, new clinical economics and universal coverage.
In this context, and considering the importance of data health, our current State-A is a situation where physician notes captured by different standards and computer systems and are at best adjoined by AI algorithms to produce, more often than not, questionable results (14). Our envisaged State-B is a situation whereby every single healthcare act is connected to other acts within a clinically coherent and continuous space of services, outcomes, data and data devices.
Also in this reality, innovation comes in two shapes, which are further compared below. In one of its shapes, sustaining innovation aims to improve existing processes and practices to make room for new ones. It does not create new markets, values or products as such. In another shape, that of disruptive innovation (15), it fuels the gradual and wide adoption of new business models and radical changes in the environment. Both are necessary.
Going forward, there are significant challenges too, including matters of scaling an AI-based or other decision tech experiment, enforcing ethical frameworks as governance frameworks, the adoption of standards for explainable and interpretable decisions, compliance with the regulatory environment to decouple innovation from sustainability - see for example the case of the Uber platform (16, 17), the complexity of the health domain and interdependence within complexity - a characteristic situation in healthcare whereby an opportunity creates opportunities for others, thus introducing further complexity.
State of innovation ecosystems
Ecosystems are biological communities of interacting organisms and the physical environment in which they coevolve capabilities (18). In their virtual reality version, or Metaverse (19), ecosystems expose the complex and connected nature of innovation, and as the Covid-19 pandemic showed, the threats, weaknesses and need for trust, regulation, meta-organisation, synergy, interoperability, solidarity, and digital foreign policy and diplomacy (20, 21). The latter represent efforts to extend the reach of innovation ecosystems as well as to enable global collaboration and coevolution, and hence the definition of global health innovation. Essentially, without these principles innovation in health cannot foster, nor be sustainable.
Connecting the elements of an innovation ecosystem are data - the connecting tissue, provided that (1) we are able to establish and maintain regulated uninterrupted data flows (blood flows to tissue), and (2) that our data can safeguard and challenge basic ethical values in technology for development (nutrients are carried to tissue and waste carried away). Without these two preconditions health innovations are essentially disconnected artefacts in an ecosystem which performs very badly under conditions of fragmentation and discontinuity. Data health is compromised and eventually so is ecosystem health.
And although the health ecosystem came under considerable and unprecedented pressure for innovation during the early stages in the Covid-19 pandemic, both of these preconditions for a functioning ecosystem were not fulfilled. Instead the pressure exposed an overall low performance bar and considerable latency from measurement to response with innovation, and back (22, 23).
At the same time, the need for primarily preventive solutions causes an explosive proliferation of innovation activity: in the US alone, start-ups in digital health raised during the first six months of 2021 up to $14.7 billion —more than the entire 2020 financing, with a total of $51.3 of funds invested into global health-tech start-ups in 2021 (24). It is not hard to imagine the ecosystem shall eventually implode.
The ecosystem reacts
The Ask Delphi research programme (24) provides some very simple demonstrations of the effect of innovation ecosystem health on AI, and in this case on machine learning in particular (25, 26).
The programme, developed by researchers at the University of Washington and the Allen Institute for Artificial Intelligence, aims to teach AI about human values, in a quest toward Artificial General Intelligence (AIG) in the sense associated with the Turing test (27). The researchers used a powerful?AI model trained to handle language by feeding on millions of sentences scraped from books and the web and gave it extra training by feeding it the consensus answers from crowd workers on Mechanical Turk (28) to ethical questions posed in Reddit forums. One of the questions found acceptable by the AI was posed as follows:
"Shooting random people with blow-darts filled with the Johnson & Johnson vaccine in order to end the pandemic".
Other researchers forewarn of an anticipated implosion in the field, setting brain health technologies as an example (29), a domain which is considered a champion in raising digital health capital. Experts warn that ballooning valuations contradict the underlying fundamental problems companies face, with hype distracting from deeper issues faced by numerous companies, such as false claims regarding benefits and poor data privacy practices, leading to a disconnect between commercially successful apps and those that have been clinically validated, with consumers and clinicians charged with the task to differentiate technologies that work, from hype.?
The innovator's predicament
To understand how innovation processes interact with the innovation ecosystem and how the latter reacts toward flawed artefact integration processes - for instance by supporting flawed or invalid reasoning, one needs to observe the multidisciplinary system dynamics which develop within health ecosystems as a result of specific innovation experimentation and development processes and stages.
As depicted in Figure-1 (adapted from 30), comparatively little is known about the ultimate disruption in the market or care environment during the early stages in product or service development, and even less about the ecosystem within which an innovation develops and functions. This situation has been named as the Innovator's Predicament following the definition of the psychology term Egocentric Predicament, coined by Ralph Barton Perry in the Journal of Philosophy (1910) to refer to the fact and predicament that we are each limited or confined to our own perceptual world, formed by essentially private experiences, thus significantly limiting our ability to coproduce and coevolve collaboratively within a common ecosystem. Given our experience is private, it is difficult to understand how genuine and thus trusted communication between two people might be possible, since both the content and symbols of any communication will be similarly private - something also known as the cartesian self (31).
In terms of digital development and innovation in health, this innovator's predicament is expressed as the inability of the innovator to view an experiment being pursued or the scaling of this experiment in terms of its integration into clinical, public health and health system practice and economies, outside their own vantage point. They can only perceive and factor-in the part of the model, platform and ecosystem which is directly relevant to their innovation and this is inherently irrelevant in clinical medicine and clinical research, certainly more so in public health, as these are inference arts and sciences which depend on denominators, continuity and integration [of inference] in order to yield results and benefits to patients. In medical AI one of the problems caused by this predicament, or one of the manifestations of the limitations caused, is bias exacerbation or the lack of longitudinal data which offer enough temporal evidence to render clinical trials reliable (32).
To undo the evident Gordian knot, it is necessary that innovation is approached from two opposite directions as shown in Figure-1: namely artefact, disruptive or business model innovation and sustaining or meta-systemic innovation, the latter of which requires the standardisation and normalisation of the measurement and information processes and tools in medicine and public health in a way that best serves both specific processes and the type of problem at hand - for instance cancer treatment management or antenatal prevention services. Then, by systematically pursuing sustaining innovation for a diverse set of problems and processes, an entire ecosystem can be modelled and metaverses developed. Welcome to the world of data ethics, macro-ethics or ecosystem ethics (33, 34).
The innovator's predicament against the backdrop of a very active investment arena such as digital health in the Covid-19 era is met by an equally daunting digital development policy makers' predicament, both of which are best depicted in terms of the current state they are producing?- a growing number of decision informing devices which interconnect using poor quality data and hence provide questionable value.
To illustrate this thought, imagine a wearable measurement device generating data which is collated outside any reliable phenotype structure and must be associated with genotype data a posteriori (posterior to care process or ground truth). Better yet consider Figure-2, which depicts the percentage of time allocated to machine learning project tasks. Two types of task are specifically charted: data curation and AI development. Deployment tasks such as validation, clinical evaluation, and explainability and interpretability testing are not included. The pie chart illustrates very clearly the impact of the ecosystem or the innovator's predicament in the average ML project, whereby 85% of effort and thus investment goes to data curation and 15% to the actual development (35). Embedding this and any other type of innovation within an competent ecosystem would render a significant efficiency improvement, ethical deployment, sustainable investment and increasing returns to scale and scope - all features of innovation within a circular digital economy.
Figure-2. Machine learning project tasks with resource allocation.
The impact of the ecosystem on innovation is also depicted in a recent McKinsey report which indicates 95% of value in German Start-ups is produced after Series-C financing, pointing to the need to establish a robust start-up ecosystem enabled by technology to deliver sustainable GDP growth (36).
Space Debris
In medicine, at the rate at which innovation artefacts are produced, usually involving either mobile health-related services or pseudo-AI services, the ecosystem resembles increasingly more the view depicted in Figure-3 of space debris around earth.
Since innovation in health and medicine are data-driven, this accelerated and underregulated development process leads to the generation and recording of debris - data of limited use and useful life, referring to the generation of pieces of clinical information which are produced during “laboratory innovation” in complete lack of harmony with the ecosystem, and which are then re-sold to the highest bidder in what has come to be referred to as Big data.
To compile and validate meaning out of such space junk, large amounts of effort and investment are required; investments which cannot be "written off" as data has to be re-curated for every single use, and which therefore significantly impact the economic viability and sustainability of investments within the current digital development ecosystem.
Figure-3. Space debris
The macro-ethics or ecosystem challenge
Until Covid-19, data sharing was very limited and, when it did take place, was very limited in scope and complexity. During the pandemic, shared data was activated under GDPR provisions for derogation from the terms of basic protection and the regulation of secondary uses of data - something called substantial?public-interest conditions.
This would have worked if a safety net of policy guidelines or policies were in place in order to define specific uses and conditions for derogation. In effect, little did take place to protect data beyond the basic consent management implemented with the EU’s General Protection Regulation (GDPR, 37), and given inadequate policies are currently in place, with few notable exceptions (38), unreliable protection processes and regulatory frameworks are being implemented in this area, which often downgrade valuable digital assets to “junk” rating once data is captured at source.
In fact, two key studies describe the demonstrated capacity of ecosystems to support strategic secondary uses of health data during the period 2020-2021 (39,40), both of which report at best mediocre results. The studies indicate the need to use data to address the multi-dimensional implications of the COVID-19 pandemic with more sophisticated products or services during the initial stages of the crisis was not met. They also suggest that governments were not fully prepared or lacked the capacity to release relevant, high-quality datasets with the speed and quality necessary that can help address the crisis. Overall, the findings demonstrate that there are several policy challenges for governments to enable data re-use with high impact. The core of such a policy is addressed in this article.
Which gives rise to the macro-ethics or ecosystem challenge, ultimately arising in data ethics (see also Toolbox-B above), to then address the problem of the space junk trap - drawing an analogy to actual space junk (41), i.e. to address the problem of siloed data with a very short useful life outside the data source environment. In fact a generation of AI applications and platforms seem to have been devised to deal with this problem alone.
Instead, a universal and long-term solution to the macro-ethics challenge and to dealing with the ecosystem's health issues, starts with enabling data sharing, which is the purpose of the GDPR, but also involves action to extend the regulatory framework to include other aspects of data protection - primarily those relating to data accuracy, i.e. the accurate interpretation of data. The idea is that extended protection that safeguards interpretative contexts, which is called semantics, and reuse (i.e. accuracy), provides a secure environment within which data can be shared with trust and thus provides for access to a higher quality and reliability data and thus ecosystem that resolves the innovators' predicament (for further information on data standards see 42).
Thus going forward action is needed to:
Toward a Circular Digital Economy
The circular digital economy is a growth model that emphasizes the extension of the product life-cycle, the reduction of the carbon footprint of digital development, and the re-use of services and goods such as data, software, infrastructure, and the transfer of know-how. In health this ecosystem is based on the extension of the semantic life of data, which is currently a deregulated niche market.
In contrast to the widely held belief that care systems and practices cause the silo phenomenon, it is our lack of policies to promote mechanisms that build into data sharing the protection of semantics that causes the silo effect, as integrated, seamless and continuous care inherently rely on data sharing.
As a result of our failure to implement a circular economy in digital development, we are currently dealing with the impact of a process and sustainability carbon footprint which is unacceptably large: a magnitude of 85% waste is not an investment opportunity.?
Going forward we need:
My data, management, integration, and trust
Distributed ledger technologies such as the blockchain have emerged through the need to implement decentralised, self-sovereign data environments and data sharing activities and provide mechanisms which are ideally suited to design and implement extended data protection policies. The blockchain paradigm:
The goal is to extend trust services toward a distributed and semantically federated blockchain of health data, including the collective digital fingerprinting of each record entry.
Benefits
The benefits of a healthy and functional innovation ecosystem in healthcare and public health are:
The Ethics of Metaverses
In its Metaverse version, the digital health innovation ecosystem shall provide sustaining innovation and macro-ethics through the following mechanisms (adapted from 43):